Home › Prompts › FLUX (Black Forest Labs) › Caption a LoRA training set for FLUX 3 Klein

Caption a LoRA training set for FLUX 3 Klein

Have a chat model write consistent, trigger-word captions for your own images before training a FLUX 3 Klein LoRA.

FLUX (Black Forest Labs)Prompt engineeringRuleStack original

Written for Any chat AI + FLUX 3 Image Klein · Updated 2026-10-17

Note: Review and test before use. Provided as is, without warranty. See Terms.

You are captioning images for a LoRA. Trigger word: [trigger]. For each image I describe or upload, write one caption: start with "[trigger]", then describe subject, pose or arrangement, setting, lighting and camera in plain sentences. Do NOT describe the style or traits the LoRA should learn ([what the LoRA should learn]); describe everything else. Max 60 words. Output as filename: caption.

Replace everything in [square brackets] with your own details.

Settings & parameters

• Train on the base checkpoint (flux-3-image-klein-base) with your usual Diffusers LoRA training script
• 15-40 varied images you own

Variations

• Style LoRA: describe content only, never the style.
• Product LoRA: vary angles, backgrounds and light.

Tips

• Only train on images you created or have rights to.